Abstract
A structural self-organized DBN (S-DBN) is proposed in this paper to improve the ability of feature learning in unsupervised training. In S-DBN, the strategy of dropout is designed for unsupervised learning to reduce inner cooperation between feature detectors. Then, the regularization-reinforced transfer function is put forward, in order to further reduce the insignificant weights, and to raise the abilities of feature learning and generation. The fast training method of contrastive divergence is designed, and backpropagation is used in supervised training. Finally, two experiments on regression and classification using MNIST show that S-DBN has better generation and faster convergence rate than other methods, in particular, in regression experiment, the proposed model beats traditional DBN by 1.50%; in image classification, the proposed model achieves smaller testing error in much less computing time.
Similar content being viewed by others
References
Prakken H, Sartor G (2015) Law and logic: a review from an argumentation perspective. Artif Intell 227:214–245
Hinton.G.E, Salakhutdinov.R.R. Reducing the dimensionality of data with neural networks. Science, 2006. 313(5786): 504–507.
LeCun Y, Bengio Y, Hinton GE (2015) Deep learning. Nature 521(7553):436–444
Hinton GE, Osindero S, Teh Y (2006) A fast learning algorithm for deep belief nets. Neural Comput 18:1527–1554
Deselaers T, Hasan S, Bender O (2009) A deep learning approach to machine transliteration. In: Proceedings of the 4th workshop on statistical machine translation, pp 233–241
Kunlei Z, Xuewen C (2014) Large-scale deep belief nets with MapReduce. Access IEEE 2:395–403
Yoshua B, Pascal L, Dan P, Hugo L (2007) Greedy layer-wise training of deep networks. In: Conference on neural information processing systems, pp 153–160
Shenhua G, Yuting Z, Kui J et al (2015) Single sample face recognition via learning deep supervised autoencoders. IEEE Trans Inform Forensics Secur 10:2108–2118
Qiao J, Pan G, Han H (2019) A regularization-reinforced DBN for digital recognition. Nat Comput 18(4):721–733
Dahl GE, Dong Y, Deng L (2011) Large vocabulary continuous speech recognition with context-dependent DBN-HMMS. In: Proceedings of IEEE international conference on acoustics, speech and signal processing, pp 4688–4691
Shenghua Z, Yan L, Bin L et al (2015) Query-oriented unsupervised multi-document summarization via deep learning model. Expert Syst Appl 42:8146–8155
Ren Z, Furao S, Jinxi Z (2014) A model with fuzzy granulation and deep belief networks for exchange rate forecasting. In: IEEE international joint conference on neural networks, pp 366–373
Zhang S, Bao Y, Zhou P, et al. (2014) Improving deep neural networks for LVCSR using dropout and shrinking structure In: 2014 IEEE international conference on acoustics, speech and signal processing, pp 6849–6853
M. Honegger (2018) Shedding light on black box machine learning algorithms: development of an axiomatic framework to assess the auality of aethods that explain individual predictions, arXiv preprint https://arXiv.org/1808.05054
Yoshua B (2009) Learning deep architectures for AI. Found Trends Mach Learn 2(1):1–127
Guangyuan P, Junfei Q, Wei C, et al. (2014) An improved RBM based on Bayesian regularization. In: IEEE international joint conference on neural networks, pp 2935–2939
Längkvist M, Karlsson L, Loutfi A (2014) A review of unsupervised feature learning and deep learning for time-series modeling. Pattern Recognit Lett 42(6):11–24
Jurgen S (2014) Deep learning in neural networks: an overview. Technical Report IDSIA-03-14, pp 1–88
Guidotti R, Monreale A, Ruggieri S, Turini F, Giannotti F, Pedreschi D (2018) A survey of methods for explaining black box models. ACM Comput Surv (CSUR) 51(5):93
Chen D, Lv J, Yi Z (2018) Graph regularized restricted Boltzmann machine. IEEE Trans Neural Netw Learn Syst 29(6):2651–2659
Yokoyama Y, Katsumata T, Yasuda M (2019) Restricted Boltzmann machine with multivalued hidden variables: a model suppressing over-fitting. Rev Socionetw Strateg 13:1–14
Jia LI, David HE, Yong QU (2019) Diagnosis of gear early pitting faults using PSO optimized deep neural network. J Northeast Univ (Natural Science) 40(7):974–979
Deepa SN, Baranilingesan I (2018) Optimized deep learning neural network predictive controller for continuous stirred tank reactor. Comput Electr Eng 71:782–797
Nair H, Tan C, Zeng M, Mengshoel OJ, Shen JP (2019) AttriNet: learning mid-level features for human activity recognition with deep belief networks. In: ACM international joint conference on pervasive and ubiquitous computing, pp 510–517
Yang H, Hu B, Pan X, Yan S, Feng Y, Zhang X (2017) Deep belief network-based drug identification using near infrared spectroscopy. J Innov Opt Health Sci 10(2):545–551
Voosen P (2017) How AI detectives are cracking open the black box of deep learning. Science 357(6346):22–28
Jun-Fei Q, Guang-yuan P, Hong-gui H (2015) The design and application of continuous deep belief network. Acta Automatica Sinica 12(41):2138–2146
Tan B, Wong A, Bizheva K (2018) Enhancement of morphological and vascular features in OCT images using a modified Bayesian residual transform. Biomed Opt Express 9(5):2394–2406
Guo J, Wang L, Hou B, Lv J, Yuan Y, Mao Q (2019) Two-phase probabilistic collaborative representation-based classification. Expert Syst Appl 133:9–20
[online] deeplearning.net. 2019
Hinton GE, Srivastava N, Krizhevsky A (2012) Improving neural networks by preventing co-adaptation of feature detectors. arXiv preprint https://arXiv.org/1207.0580
Acknowledgments
This work was funded by Beijing Postdoctoral Science Foundation under Grant ZZ-2019-65, Chaoyang District Postdoctoral Science Foundation under Grant 2019ZZ-45 and Beijing Municipal Education Commission under Grant KM201811232016.
Funding
This work was funded by Beijing Postdoctoral Science Foundation under Grant ZZ-2019-65, Chaoyang District Postdoctoral Science Foundation under Grant 2019ZZ-45 and Beijing Municipal Education Commission under Grant KM201811232016.
Author information
Authors and Affiliations
Contributions
Conceptualization, Qili Chen and Guangyuan Pan; methodology, Qili Chen and Guangyuan Pan; software, Qili Chen and Guangyuan Pan; validation, Qili Chen jinjin Jiang and Guangyuan Pan; formal analysis, Guangyuan Pan; investigation, Guangyuan Pan; resources, Guangyuan Pan; data curation, Guangyuan Pan; writing—original draft preparation, Qili Chen jinjin Jiang and Guangyuan Pan; writing—review and editing, Qili Chen jinjin Jiang and Guangyuan Pan; visualization, Qili Chen jinjin Jiang and Guangyuan Pan; supervision, Guangyuan Pan; project administration, Guangyuan Pan; funding acquisition, Qili Chen. All authors have read and agreed to the published version of the manuscript.
Corresponding author
Ethics declarations
Conflict of interest
All the authors declare no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Chen, Q., Pan, G. A structure-self-organizing DBN for image recognition. Neural Comput & Applic 33, 877–886 (2021). https://doi.org/10.1007/s00521-020-05262-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00521-020-05262-2